Errors & Config

Shared configuration and state types: session and user state enums, speech and VAD event types, and the config objects for end-of-utterance, interruption, and realtime behavior.

UserState

States describing the user’s current speech activity in a session.

class UserState(Enum):
    IDLE = 'idle'
    SPEAKING = 'speaking'
    LISTENING = 'listening'

AgentState

Lifecycle and activity states of the agent during a session.

class AgentState(Enum):
    STARTING = 'starting'
    IDLE = 'idle'
    SPEAKING = 'speaking'
    LISTENING = 'listening'
    THINKING = 'thinking'
    CLOSING = 'closing'

SpeechEventType

Types of speech-related events emitted during transcription.

class SpeechEventType(Enum):
    START = 'start_of_speech'
    INTERIM = 'interim_transcript'
    PREFLIGHT = 'preflight_transcript'
    FINAL = 'final_transcript'
    END = 'end_of_speech'

VADEventType

Types of voice activity detection events.

class VADEventType(Enum):
    START_OF_SPEECH = 'start_of_speech'
    END_OF_SPEECH = 'end_of_speech'

EOUConfig

Configuration for end-of-utterance (EOU) detection.

Fields

mode

Literal['ADAPTIVE', 'DEFAULT']

default:"DEFAULT"

Detection strategy, either 'ADAPTIVE' or 'DEFAULT'.

min_max_speech_wait_timeout

List[float] | Tuple[float, float]

default:"…"

Minimum and maximum wait times, in seconds, before concluding that the user has finished speaking.

InterruptConfig

Configuration controlling how user interruptions of the agent are detected and handled.

Fields

mode

Literal['VAD_ONLY', 'STT_ONLY', 'HYBRID']

default:"HYBRID"

Detection strategy, one of 'VAD_ONLY', 'STT_ONLY' or 'HYBRID'.

interrupt_min_duration

float

default:"0.5"

Minimum speech duration, in seconds, required to count as an interruption.

interrupt_min_words

int

default:"2"

Minimum number of recognized words required to count as an interruption.

interrupt_min_confidence

float

default:"0.0"

Minimum transcription confidence required to count as an interruption.

false_interrupt_pause_duration

float

default:"2.0"

Pause duration, in seconds, used when an interruption is later judged to be false.

resume_on_false_interrupt

bool

default:"True"

Whether to resume the agent’s response after a false interruption.

false_interrupt_pause_duration_ms

int

default:"2000"

Pause duration in milliseconds applied on a false interruption.

interrupt_fade_duration

float

default:"0.0"

Audio fade-out duration, in seconds, applied when an interruption occurs.

interrupt_fade_duration_ms

int

default:"400"

Audio fade-out duration in milliseconds applied when an interruption occurs.

RealtimeConfig

Configuration for a realtime model session.

Fields

mode

Literal['full_s2s', 'hybrid_stt', 'hybrid_tts', 'llm_only'] | None

The realtime operating mode, one of 'full_s2s', 'hybrid_stt', 'hybrid_tts' or 'llm_only', or None to use the default.

response_modalities

List[str] | None

The output modalities the model should produce, or None for the default.

Python SDK

Core

Speech-to-text

LLM

Text-to-speech

Realtime models

Avatars

Turn detection

Voice activity detection

Noise cancellation

UserState

AgentState

SpeechEventType

VADEventType

EOUConfig

Fields

InterruptConfig

Fields

RealtimeConfig

Fields

​UserState

​AgentState

​SpeechEventType

​VADEventType

​EOUConfig

​Fields

​InterruptConfig

​Fields

​RealtimeConfig

​Fields

UserState

AgentState

SpeechEventType

VADEventType

EOUConfig

Fields

InterruptConfig

Fields

RealtimeConfig

Fields